Mayo Clinic is going public with own COVID-19 modeling

Mayo Clinic is going public with a modeling approach to predicting COVID-19 trends that its leaders said has effectively guided their hospital preparations since the start of the pandemic.

The Rochester-based health system is unveiling an online tracking tool Thursday that provides recent COVID-19 case data for U.S. counties, as well as reputable advice on how to prevent infections with the coronavirus that causes the disease. The tool on the mayoclinic.org site will be updated this fall with county-level modeling forecasts for pandemic activity.

"There are a lot of websites — including the Johns Hopkins website, IHME, Google, CDC — and a lot of their predictions they're doing about cases and hospitalizations are at the state level," said Dr. Henry Ting, Mayo's chief value officer, who led the development of the tracker. "We thought there was a need for people to have information about cases … at the county level."

Modeling has been a key but controversial component of national and state responses to the pandemic. The University of Minnesota and the Minnesota Department of Health created models that forecast how changes such as business closures and other restrictions would reduce viral spread.

Gov. Tim Walz factored that trend data in his decision to enact a state shutdown for 51 days this spring, but received criticism from political opponents because the models overestimated COVID deaths.

State Health Commissioner Jan Malcolm said on Wednesday that she couldn't disclose specifics, but that U researchers had revised methodology for a new round of predictive modeling on COVID-19 trends.

"It's being worked on ... with some different updates to the model, different improvements to parameters and assumptions based on more Minnesota data," she said.

IHME, or the Institute for Health Metrics and Evaluation, in Washington state, has been predicting death tolls since the start of the pandemic based partly on states' response strategies to slow the spread of the virus. It currently predicts a total of 4,990 COVID-19 deaths in Minnesota by year's end, but only 2,372 deaths if the state maintains a universal mask mandate and 95% of residents follow it.

Mayo leaders had used modeling predictions internally to divert resources to its hospitals in Minnesota and other states, and to advise the state on COVID-19 hot spots where more diagnostic tests and medical resources might be needed, but hadn't shared the data publicly.

One of the first public acknowledgments of its modeling efforts came in mid-May, when Mayo chief executive Dr. Gianrico Farrugia suggested that Minnesota's shutdown efforts had reduced the peak amount of virus transmission — even though some models suggested that the peak was only delayed.

"Mayo Clinic's predictive modeling shows that the anticipated COVID-19 very high peak has been diminished," he wrote at the time, "and we will experience varied levels of infection rates in the coming months until an effective vaccine is available."

Ting said he believes the Mayo model works better than others because it is fluid and doesn't get hindered by static mathematical assumptions that don't get updated frequently enough.

Mayo's model "takes into account that there is uncertainty in those parameters," Ting said. "So we're allowing those parameters to change on a daily basis and we're running our models every day on a dynamic basis."

Forecasts on the public tracking tool will be looking only a week or two ahead, whereas internal modeling has looked farther out.

Ting said the tracking tool will offer much more than just modeling predictions. Mixed public health and media messages on subjects such as masks have created an opportunity for Mayo to weigh in with its expertise.

"We can provide Mayo-vetted information about what we're learning about COVID diagnosis, testing, treatment and vaccines," he said, "and hopefully for that to be nonpolitical and mostly science-based and data-based."

Mayo is not exploring the use of mobile phone apps to inform people through mobility data when they have spent time close enough to others diagnosed with COVID-19 to be at risk for exposure.

Updates to Google and Apple operating platforms have made this easier to accomplish, and countries in Asia and Europe have created versions to track the spread of COVID-19 and to alert people to infection risks. However, Ting said many in the U.S. would likely view that as an invasion of privacy.

HealthPartners this spring offered a SafeDistance crowdsourcing app that encouraged users to voluntarily submit their anonymous illness information so that users in their surrounding neighborhoods would know the level of risk.

North Dakota's Care19 Alert app used Bluetooth technology to notify participants when they had been close to other users of the app who later tested positive for COVID-19.